It's nothing to do with encoding. If the text includes the characters with code points 2 and 7, then they can't be included in an XML 1.0 document. (It sounds as if your conversion tool ignores characters that aren't invalid in XML, other tools might throw an error, or find some other way of representing them for example as a processing instruction <?char 2?>).
XML 1.1 allows these code points, but XML 1.1 isn't widely supported.